CN111309930A - Medical knowledge graph entity alignment method based on representation learning - Google Patents

Medical knowledge graph entity alignment method based on representation learning Download PDF

Info

Publication number
CN111309930A
CN111309930A CN202010151549.6A CN202010151549A CN111309930A CN 111309930 A CN111309930 A CN 111309930A CN 202010151549 A CN202010151549 A CN 202010151549A CN 111309930 A CN111309930 A CN 111309930A
Authority
CN
China
Prior art keywords
entity
entities
aligned
medical knowledge
recommending
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010151549.6A
Other languages
Chinese (zh)
Other versions
CN111309930B (en
Inventor
滕飞
钟文
许强
李天瑞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwest Jiaotong University
Original Assignee
Southwest Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southwest Jiaotong University filed Critical Southwest Jiaotong University
Priority to CN202010151549.6A priority Critical patent/CN111309930B/en
Publication of CN111309930A publication Critical patent/CN111309930A/en
Application granted granted Critical
Publication of CN111309930B publication Critical patent/CN111309930B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Machine Translation (AREA)

Abstract

The invention provides a medical knowledge map entity alignment method based on representation learning, wherein complete character strings of entities are matched, and common entities are removed to obtain entities to be aligned; training triples in the two knowledge maps through a knowledge representation learning model to obtain an embedded vector of each entity; calculating the semantic distance between the entity to be aligned and the standard entity, and finding out the first ten recommended entities with the highest semantic distance value; removing candidate entities with completely different root words by constructing a root word set of medical knowledge data; judging whether the entity to be aligned and the recommending entity are in the same department and part, and removing unreliable results in the recommending entity; and marking the finally obtained entity to be aligned and the recommending entity, finding out the entity which is the same as the entity to be aligned in the recommending entity, and finishing entity alignment. On the basis of expression learning, the root word and the rule are added to screen candidate entities according to the characteristics of knowledge in the medical field, and a more accurate entity alignment result is obtained.

Description

Medical knowledge graph entity alignment method based on representation learning
Technical Field
The invention belongs to the technical field of natural language processing, and particularly relates to a medical knowledge graph entity alignment method based on representation learning.
Background
With the development of the internet, the search demand of internet medical science popularization is increased rapidly, and the existing medical science popularization websites on the internet are various, such as a disease department network, a 39 health network, a medical inquiry and drug search network and the like. Although these websites contain rich medical knowledge, website navigation is too professional to allow ordinary users to quickly find needed content, and the website navigation lacks pertinence and cannot give targeted answers according to different questions of the users, so that the task of constructing intelligent medical treatment based on knowledge graph technology is not slow. The knowledge graph is a large-scale semantic network, expresses knowledge in the form of triples (head entities, relations and tail entities), and is widely applied to the fields of information retrieval, intelligent question answering, recommendation systems and the like, while the medical knowledge graph has numerous data sources and usually contains a large amount of overlapped triples of information. If medical knowledge-maps of different data sources are to be used simultaneously, the entities must be aligned to determine whether different entities in the difference data sources point to the same thing in the real world. Since the precision requirement of the medical field for the entity alignment result is very high, a great challenge is brought to the implementation of the alignment work in the field.
The invention provides a knowledge fusion method based on multi-source data in the field of general knowledge maps, which is based on the entity name and entity attribute to perform block aggregation on entities, takes entities from different sources in the same block as a candidate entity pair to reduce the calculation complexity, then adopts an entity alignment algorithm to calculate the similarity between the entities, and if the similarity is greater than a preset threshold value, the two entities are considered to point to the same entity, and finally obtains equivalent links of all the entities between different data sources; and generating an entity alignment seed set according to a preset frequency selection method, then generating a joint embedding space corresponding to the triple through a relation triple and attribute triple joint embedding module, iteratively training and selecting a pair of entities with the minimum semantic distance to form an entity pair, and finally adding the entity pair meeting a preset distance threshold value to the entity alignment seed set for updating until the entity pair meeting the preset distance threshold value does not exist.
However, the method depends on a preset threshold, the selection of the threshold has no fixed standard, the threshold can be estimated only by experience and experiments, the result of entity alignment is decisively influenced, and the reliability of the alignment result cannot be judged, so that the method is difficult to be applied to the medical field with high precision requirement. Because the number of knowledge graph entities is huge, all entities with the same direction are ensured to be found out in the alignment process, and the alignment workload is also reduced, on the basis of expression learning, a root word and a rule are added to screen candidate entities according to the characteristics of knowledge in the medical field, so that a more accurate alignment result is obtained, and the root word is used as a substring with representative significance in the entities and can reflect important characteristics in the medical entities.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides the medical knowledge graph entity alignment method based on representation learning, which solves the problems that threshold selection is fuzzy and a model structure is single in medical knowledge graph entity alignment, so that the reliability of an alignment result cannot be ensured.
In order to achieve the above purpose, the invention adopts the technical scheme that:
the scheme provides a medical knowledge graph entity alignment method based on representation learning, which comprises the following steps:
s1, removing the standard medical knowledge map K1And medical knowledge map K to be aligned2The common entity in the Chinese medical science knowledge map K is obtained1Of residual entity E'1And medical knowledge map K to be aligned2Of residual entity E'2；
S2, respectively aligning the standard medical knowledge maps K by the knowledge representation learning model1And medical knowledge map K to be aligned2All triplets T in1And T2Training to obtain residual entity E'1The embedded vector and remaining entity E 'of each entity'2The embedded vector of each entity in the set;
s3, calculating by utilizing a cosine similarity function to obtain a residual entity E'1The embedded vector and remaining entity E 'of each entity'2The similarity value of the embedded vector of each entity is obtained, and the residual entity E 'is obtained according to the similarity value'2Of each entity to be aligned with a remaining entity E'1The first ten recommended entities with the highest similarity value of all the entities in the group;
s4, according to the standard medical knowledge map K1All entities E of1And medical knowledge map K to be aligned2All entities E in2Constructing a word root set and acquiring a residual entity E 'by utilizing the word root set'2The root of the entity to be aligned and the roots of the first ten recommended entities are removed, and the recommended entities with different roots are removed;
s5, judging the residual entity E'2Whether the department or part where the entity to be aligned is located is the same as the recommended entity with the different root deleted { e1,e2,...,emThe departments or parts where the entities are located are the same, if yes, the recommending entity is reserved, otherwise, the recommending entity is deleted, and a new recommending entity { e ] is obtained according to the judgment result1,e2,...,elM represents the number of recommended entities after different word roots are deleted, and l represents the number of recommended entities after different departments or parts are deleted;
s6, mixing the rest entity E'2The entity to be aligned with the new recommending entity { e } in step S51,e2,...,elMark and find out new recommending entity { e }1,e2,...,elAnd residual entity E'2Wherein the entities to be aligned point to the same entity, thereby completing the alignment of the medical knowledge map entities.
Further, the expression of the similarity value in step S3 is as follows:
Figure BDA0002402608850000031
wherein,
Figure BDA0002402608850000032
the value of the degree of similarity is represented,
Figure BDA0002402608850000033
represents residual entity E'1The embedded vector of each of the entities in the set,
Figure BDA0002402608850000041
represents residual entity E'2The embedded vector of each of the entities in the set,
Figure BDA0002402608850000042
and
Figure BDA0002402608850000043
respectively representing embedded vectors
Figure BDA0002402608850000044
And
Figure BDA0002402608850000045
the die of (1).
Still further, the step S4 includes the steps of:
s401, according to the standard medical knowledge map K1All entities E of1And medical knowledge map K to be aligned2All entities E in2Constructing dictionary W, W ═ W1,W2,...,WnAnd to WiPerforming substring resolution, wherein WiDenotes a word in the dictionary, i 1,21And E2Is the sum of the number of entities;
s402, placing the analyzed substrings in the whole dictionary W respectively for frequency statistics, and selecting the substrings 3 before the frequency as roots;
s403, judging whether the whole dictionary W is traversed or not, if so, performing deduplication processing on the selected root word to obtain a root word set WETo make a hand in hand withStep S404 is entered, otherwise, the step S402 is returned to;
s404, respectively aligning residual entities E'2The entity to be aligned with the remaining entity E 'corresponding to the entity to be aligned'1Performing substring analysis on the first ten recommended entities, and judging whether the analyzed substrings exist in the root set WEIf yes, respectively obtaining residual entity E'1Substring set of ten first recommended entities { W }e11,We12,...,We110And remaining entity E'2In the substring set W of the entities to be alignede2And step S405 is proceeded, otherwise, step S404 is repeated until the residual entity E 'is traversed'2All substrings and E'1Substrings of the first ten recommended entities;
s405, judging a substring set { We11,We12,...,We110W and the substring sete2If yes, the recommending entity for generating the substring set is reserved, otherwise, the recommending entity is deleted, the recommending entities with different root words are removed, and a new recommending entity { e } is obtained according to the judgment result1,e2,...,emAnd step S5 is entered, wherein m represents the number of recommended entities without the same root.
Still further, the root word in step S402 further includes:
and selecting the substring with the longest length as the root word when the frequency numbers are the same.
Still further, step S6 is specifically:
the remaining entity E'2Wherein each entity to be aligned is in a remaining entity E'1Recommending entity { e } in (1)1,e2,...,elMark and find out the recommended entity { e }1,e2,...,elThe entities in the alignment point to the same entity as the entity to be aligned, resulting in an entity pair (e)sim1,esim2) Thereby completing the entity alignment of the medical knowledge map, wherein esim1Is residual entity E'1Recommending entity of (1), esim2Is residual entity E'2To be aligned.
The invention has the beneficial effects that:
(1) the method recommends the candidate entity through the knowledge representation learning model, and can solve the problems of insufficient feature learning and poor alignment effect caused by single model;
(2) the method fully utilizes the characteristics of medical knowledge, constructs meaningful roots, reflects important characteristics in medical entities, ensures that all entities with the same orientation are found out, and also reduces the workload of entity alignment;
(3) according to the invention, according to the special rules contained in the medical knowledge, the part of the entity and the department can judge whether the entity points to the same entity, thereby further reducing the number of recommendation results and ensuring the accuracy;
(4) the invention adopts the marking method to obtain the aligned correct entity, in the medical field, the correctness of the alignment result is related to the reliability of the medical knowledge graph, the workload of marking is greatly reduced, and the quality of the medical knowledge graph is ensured.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
Examples
Assuming that a standard medical knowledge-map K exists1＝{E1,R1,T1},E1Is K1All entities in (1), R1Is K1All relationships in (1), T1Is K1All triples in (1); medical knowledge graph K to be aligned2＝{E2,R2,T2},E2Is K2All entities in (1), R2Is K2All relationships in (1), T2Is K2All triples in (a). The final aim is to convert K2All entities E in2And K1All entities E in1Alignment is performed to find entities pointing to the same.
As shown in fig. 1, the invention discloses a medical knowledge-graph entity alignment method based on representation learning, which is implemented as follows:
s1, removing the standard medical knowledge map K1And medical knowledge map K to be aligned2The common entity in the Chinese medical science knowledge map K is obtained1Of residual entity E'1And medical knowledge map K to be aligned2Of residual entity E'2。
In this embodiment, the complete strings of entities are matched to find out the completely same entities in the two knowledge graphs, i.e. the common entity Esim. For aligned knowledge graph K1The remaining entity is E'1＝E1-EsimFor knowledge graph K to be aligned2The remaining entity is E'2＝E2-EsimThen the final purpose is converted to E'2Each entity of (a) and E'1All entities in (a) are aligned.
S2, respectively aligning the standard medical knowledge maps K by the knowledge representation learning model1And medical knowledge map K to be aligned2All triplets T in1And T2Training to obtain residual entity E'1The embedded vector and remaining entity E 'of each entity'2The embedded vector of each entity in the set;
in the present embodiment, the learning model pair K is expressed by knowledge1And K2All triplets T in (2)1And T2Training is carried out until the loss function of the knowledge representation learning model is smaller than a set value, so that E 'is obtained'2And E'1The embedded vector of each of the entities in the set,
Figure BDA0002402608850000061
is E'1Embedding of each entity inThe vector of the vector is then calculated,
Figure BDA0002402608850000062
is E'2The embedded vector of each entity.
S3, calculating by utilizing a cosine similarity function to obtain a residual entity E'1The embedded vector and remaining entity E 'of each entity'2The similarity value of the embedded vector of each entity is obtained, and the residual entity E 'is obtained according to the similarity value'2Of each entity to be aligned with a remaining entity E'1The first ten recommended entities with the highest similarity value of all the entities in the group; the expression for the similarity value is as follows:
Figure BDA0002402608850000071
wherein,
Figure BDA0002402608850000072
the value of the degree of similarity is represented,
Figure BDA0002402608850000073
represents residual entity E'1The embedded vector of each of the entities in the set,
Figure BDA0002402608850000074
represents residual entity E'2The embedded vector of each of the entities in the set,
Figure BDA0002402608850000075
and
Figure BDA0002402608850000076
respectively representing embedded vectors
Figure BDA0002402608850000077
And
Figure BDA0002402608850000078
die of
S4, according to the standard medical knowledge map K1All entities E of1And is to be alignedMedical knowledge map K2All entities E in2Constructing a word root set and acquiring a residual entity E 'by utilizing the word root set'2The root of the entity to be aligned and the roots of the first ten recommended entities are removed, and the recommended entities with different roots are removed; the realization method comprises the following steps:
s401, according to the standard medical knowledge map K1All entities E of1And medical knowledge map K to be aligned2All entities E in2Constructing dictionary W, W ═ W1,W2,...,WnAnd to WiPerforming substring resolution, wherein WiDenotes a word in the dictionary, i 1,21And E2Is the sum of the number of entities;
in this embodiment, let dictionary W be E1And E2All entities in the dictionary have n words, n is E1And E2Is the sum of the number of entities, W ═ W1,W2,...,WnTo Wi(i ═ 1, 2.., n) perform substring parsing. In the medical data, L-1+ L-2+. eta. +1 substrings exist in an entity with the length of L, at least one of the substrings is a root, the number of words for the root is not less than 2, and the higher the occurrence frequency of the substring is, the more likely the substring is to be a root for an entity. Taking the entity of "anterior patellar capsulitis" as an example, the substrings include "anterior patellar", "anterior capsule", "capsulitis", "anterior patellar capsule", "anterior capsulitis" and "anterior patellar capsulitis".
S402, placing the analyzed substrings in the whole dictionary W respectively for frequency statistics, and selecting the substrings 3 before the frequency as roots; and selecting the substring with the longest length as the root word when the frequency numbers are the same.
In this embodiment, after "anterior patellar bursitis" is disassembled, its substrings are respectively placed in the whole dictionary W for frequency statistics. And selecting the substrings with the frequency numbers of 3 before as the root words, and preferentially selecting the substrings with longer lengths if the substrings with the same frequency numbers but longer lengths exist, so that the substrings which are possibly the root words are added into the root word set.
S403, judging whether the whole dictionary W is traversed or not, if so, comparingThe selected root words are subjected to duplication elimination processing to obtain a root word set WEAnd go to step S404, otherwise, return to step S402;
s404, respectively aligning residual entities E'2The entity to be aligned with the remaining entity E 'corresponding to the entity to be aligned'1Performing substring analysis on the first ten recommended entities, and judging whether the analyzed substrings exist in the root set WEIf yes, respectively obtaining residual entity E'1Substring set of ten first recommended entities { W }e11,We12,...,We110And remaining entity E'2In the substring set W of the entities to be alignede2And step S405 is proceeded, otherwise, step S404 is repeated until the residual entity E 'is traversed'2All substrings and E'1Substrings of the first ten recommended entities;
s405, judging a substring set { We11,We12,...,We110W and the substring sete2If yes, the recommending entity for generating the substring set is reserved, otherwise, the recommending entity is deleted, the recommending entities with different root words are removed, and a new recommending entity { e } is obtained according to the judgment result1,e2,...,emFourthly, the step S5 is carried out, wherein m represents the number of recommended entities after different root words are deleted;
s5, judging the residual entity E'2Whether the department or part where the entity to be aligned is located is the same as the recommended entity with the different root deleted { e1,e2,...,emThe departments or parts where the entities are located are the same, if yes, the recommending entity is reserved, otherwise, the recommending entity is deleted, and a new recommending entity { e ] is obtained according to the judgment result1,e2,...,elM represents the number of recommended entities without the same root, and l represents the number of recommended entities with different departments or parts deleted;
s6, mixing the rest entity E'2Marking the entity to be aligned with the new recommending entity in the step S5, and finding out a new recommending entity and a residual entity E'2The entities to be aligned point to the same entity, so that the alignment of the medical knowledge graph entities is completed, and the method specifically comprises the following steps:
the remaining entity E'2Wherein each entity to be aligned is in a remaining entity E'1Recommending entity { e } in (1)1,e2,...,elMark and find out the recommended entity { e }1,e2,...,elThe entities in the alignment point to the same entity as the entity to be aligned, resulting in an entity pair (e)sim1,esim2) Thereby completing the entity alignment of the medical knowledge map, wherein esim1Is residual entity E'1Recommending entity of (1), esim2Is residual entity E'2To be aligned.
Through the design, on the basis of expression learning, according to the characteristics of knowledge in the medical field, the root of a word and rules (judgment on whether departments or parts are the same) are added to screen candidate entities, so that a more accurate entity alignment result is obtained, and the problem that the reliability of the alignment result cannot be ensured due to fuzzy threshold selection and single model structure in the entity alignment of the medical knowledge map is solved.

Claims (5)

1. A medical knowledge-graph entity alignment method based on representation learning is characterized by comprising the following steps:
s1, removing the standard medical knowledge map K1And medical knowledge map K to be aligned2The common entity in the Chinese medical science knowledge map K is obtained1Of residual entity E'1And medical knowledge map K to be aligned2Of residual entity E'2；
S2, respectively aligning the standard medical knowledge maps K by the knowledge representation learning model1And medical knowledge map K to be aligned2All triplets T in1And T2Training to obtain residual entity E'1The embedded vector and remaining entity E 'of each entity'2The embedded vector of each entity in the set;
s3, calculating by utilizing a cosine similarity function to obtain a residual entity E'1The embedded vector and remaining entity E 'of each entity'2The similarity value of the embedded vector of each entity in the tree is obtained, and the residue is obtained according to the similarity valueEntity E'2Of each entity to be aligned with a remaining entity E'1The first ten recommended entities with the highest similarity value of all the entities in the group;
s4, according to the standard medical knowledge map K1All entities E of1And medical knowledge map K to be aligned2All entities E in2Constructing a word root set and acquiring a residual entity E 'by utilizing the word root set'2The root of the entity to be aligned and the roots of the first ten recommended entities are removed, and the recommended entities with different roots are removed;
s5, judging the residual entity E'2Whether the department or part where the entity to be aligned is located is the same as the recommended entity with the different root deleted { e1,e2,...,emThe departments or parts where the entities are located are the same, if yes, the recommending entity is reserved, otherwise, the recommending entity is deleted, and a new recommending entity { e ] is obtained according to the judgment result1,e2,...,elM represents the number of recommended entities after different word roots are deleted, and l represents the number of recommended entities after different departments or parts are deleted;
s6, mixing the rest entity E'2The entity to be aligned with the new recommending entity { e } in step S51,e2,...,elMark and find out new recommending entity { e }1,e2,...,elAnd residual entity E'2Wherein the entities to be aligned point to the same entity, thereby completing the alignment of the medical knowledge map entities.
2. The representation learning-based medical knowledge-graph entity alignment method of claim 1, wherein the similarity values in step S3 are expressed as follows:
Figure FDA0002402608840000021
wherein,
Figure FDA0002402608840000022
the value of the degree of similarity is represented,
Figure FDA0002402608840000023
represents residual entity E'1The embedded vector of each of the entities in the set,
Figure FDA0002402608840000024
represents residual entity E'2The embedded vector of each of the entities in the set,
Figure FDA0002402608840000025
and
Figure FDA0002402608840000026
respectively representing embedded vectors
Figure FDA0002402608840000027
And
Figure FDA0002402608840000028
the die of (1).
3. The representation learning-based medical knowledge-graph entity alignment method of claim 1, wherein said step S4 comprises the steps of:
s401, according to the standard medical knowledge map K1All entities E of1And medical knowledge map K to be aligned2All entities E in2Constructing dictionary W, W ═ W1,W2,...,WnAnd to WiPerforming substring resolution, wherein WiDenotes a word in the dictionary, i 1,21And E2Is the sum of the number of entities;
s402, placing the analyzed substrings in the whole dictionary W respectively for frequency statistics, and selecting the substrings 3 before the frequency as roots;
s403, judging whether the whole dictionary W is traversed or not, if so, performing deduplication processing on the selected root word to obtain a root word set WEAnd go to step S404, otherwise, return to step S402;
s404, respectively aligning residual entities E'2The entity to be aligned with the remaining entity E 'corresponding to the entity to be aligned'1Performing substring analysis on the first ten recommended entities, and judging whether the analyzed substrings exist in the root set WEIf yes, respectively obtaining residual entity E'1Substring set of ten first recommended entities { W }e11,We12,...,We110And remaining entity E'2In the substring set W of the entities to be alignede2And step S405 is proceeded, otherwise, step S404 is repeated until the residual entity E 'is traversed'2All substrings and E'1Substrings of the first ten recommended entities;
s405, judging a substring set { We11,We12,...,We110W and the substring sete2If yes, the recommending entity for generating the substring set is reserved, otherwise, the recommending entity is deleted, the recommending entities with different root words are removed, and a new recommending entity { e } is obtained according to the judgment result1,e2,...,emAnd step S5 is entered, wherein m represents the number of recommended entities without the same root.
4. The representation learning-based medical knowledge-graph entity alignment method of claim 3, wherein the root word in step S402 further comprises:
and selecting the substring with the longest length as the root word when the frequency numbers are the same.
5. The representation learning-based medical knowledge-graph entity alignment method according to claim 1, wherein the step S6 is specifically:
the remaining entity E'2Wherein each entity to be aligned is in a remaining entity E'1Recommending entity { e } in (1)1,e2,...,elMark and find out the recommended entity { e }1,e2,...,elThe entities in the alignment point to the same entity as the entity to be aligned, resulting in an entity pair (e)sim1,esim2) To thereby complete the medical knowledge mapEntity alignment, wherein esim1Is residual entity E'1Recommending entity of (1), esim2Is residual entity E'2To be aligned.
CN202010151549.6A 2020-03-06 2020-03-06 Medical knowledge graph entity alignment method based on representation learning Active CN111309930B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010151549.6A CN111309930B (en) 2020-03-06 2020-03-06 Medical knowledge graph entity alignment method based on representation learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010151549.6A CN111309930B (en) 2020-03-06 2020-03-06 Medical knowledge graph entity alignment method based on representation learning

Publications (2)

Publication Number Publication Date
CN111309930A true CN111309930A (en) 2020-06-19
CN111309930B CN111309930B (en) 2023-02-28

Family

ID=71147280

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010151549.6A Active CN111309930B (en) 2020-03-06 2020-03-06 Medical knowledge graph entity alignment method based on representation learning

Country Status (1)

Country Link
CN (1) CN111309930B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112966124A (en) * 2021-05-18 2021-06-15 腾讯科技（深圳）有限公司 Training method, alignment method, device and equipment of knowledge graph alignment model
CN113420161A (en) * 2021-06-24 2021-09-21 平安科技（深圳）有限公司 Node text fusion method and device, computer equipment and storage medium
CN118193757A (en) * 2024-05-17 2024-06-14 之江实验室 Task execution method and device, storage medium and electronic equipment

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180218015A1 (en) * 2013-03-15 2018-08-02 Neura Labs Corp. Intelligent internet system with adaptive user interface providing one-step access to knowledge
CN109378053A (en) * 2018-11-30 2019-02-22 安徽影联云享医疗科技有限公司 A kind of knowledge mapping construction method for medical image
CN110188206A (en) * 2019-05-08 2019-08-30 北京邮电大学 Collaboration iterative joint entity alignment schemes and device based on translation model
CN110852895A (en) * 2019-11-13 2020-02-28 易联众信息技术股份有限公司 Medical security cheat-insurance behavior discovery method based on knowledge graph

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180218015A1 (en) * 2013-03-15 2018-08-02 Neura Labs Corp. Intelligent internet system with adaptive user interface providing one-step access to knowledge
CN109378053A (en) * 2018-11-30 2019-02-22 安徽影联云享医疗科技有限公司 A kind of knowledge mapping construction method for medical image
CN110188206A (en) * 2019-05-08 2019-08-30 北京邮电大学 Collaboration iterative joint entity alignment schemes and device based on translation model
CN110852895A (en) * 2019-11-13 2020-02-28 易联众信息技术股份有限公司 Medical security cheat-insurance behavior discovery method based on knowledge graph

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
LIU XIAOXUE: "Review and Trend Analysis of Knowledge Graphs for Crop Pest and Diseases", 《IEEE》 *
吴运兵等: "知识图谱学习和推理研究进展", 《小型微型计算机***》 *
朱继召等: "表示学习知识图谱的实体对齐算法", 《东北大学学报(自然科学版)》 *
滕飞: "复杂网络的重叠社区发现并行算法", 《西南交通大学学报》 *
苏佳林等: "自适应属性选择的实体对齐方法", 《山东大学学报(工学版)》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112966124A (en) * 2021-05-18 2021-06-15 腾讯科技（深圳）有限公司 Training method, alignment method, device and equipment of knowledge graph alignment model
CN112966124B (en) * 2021-05-18 2021-07-30 腾讯科技（深圳）有限公司 Training method, alignment method, device and equipment of knowledge graph alignment model
CN113420161A (en) * 2021-06-24 2021-09-21 平安科技（深圳）有限公司 Node text fusion method and device, computer equipment and storage medium
CN118193757A (en) * 2024-05-17 2024-06-14 之江实验室 Task execution method and device, storage medium and electronic equipment

Also Published As

Publication number Publication date
CN111309930B (en) 2023-02-28

Similar Documents

Publication Publication Date Title
CN110399457B (en) Intelligent question answering method and system
CN110413732B (en) Knowledge searching method for software defect knowledge
JP6901816B2 (en) Entity-related data generation methods, devices, devices, and storage media
WO2018040899A1 (en) Error correction method and device for search term
CN109726298B (en) Knowledge graph construction method, system, terminal and medium suitable for scientific and technical literature
CN112380325B (en) Knowledge graph question-answering system based on joint knowledge embedded model and fact memory network
CN112434169A (en) Knowledge graph construction method and system and computer equipment
CN104866593A (en) Database searching method based on knowledge graph
CN102663129A (en) Medical field deep question and answer method and medical retrieval system
WO2020010834A1 (en) Faq question and answer library generalization method, apparatus, and device
CN110765277B (en) Knowledge-graph-based mobile terminal online equipment fault diagnosis method
CN110175585B (en) Automatic correcting system and method for simple answer questions
CN110321437B (en) Corpus data processing method and device, electronic equipment and medium
CN111026815B (en) Entity pair specific relation extraction method based on user-assisted correction
CN115599899B (en) Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph
CN113590784A (en) Triple information extraction method and device, electronic equipment and storage medium
CN111309930B (en) Medical knowledge graph entity alignment method based on representation learning
US20090234852A1 (en) Sub-linear approximate string match
CN113742446A (en) Knowledge graph question-answering method and system based on path sorting
CN115658846A (en) Intelligent search method and device suitable for open-source software supply chain
CN112395425A (en) Data processing method and device, computer equipment and readable storage medium
CN113254671B (en) Atlas optimization method, device, equipment and medium based on query analysis
CN114579606B (en) Pre-training model data processing method, electronic device and computer storage medium
CN115114417A (en) Automatic scoring method and device for Chinese blank filling questions
CN114090722A (en) Method and device for automatically completing query content

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant